In response to the rapid growth in payments for post-acute care (PAC) services, Congress enacted Medicare reforms as part of the Balanced Budget Act of 1997 (BBA) for each PAC service. The reforms mandated a series of separate case-mix adjusted prospective payment systems, each with its own implementation timeline. In addition to the overall effects, the BBA's effects on PAC use varied substantially across geographic areas. For example, in the case of skilled nursing facility use from 1998 to 2000, the average change relative to 1996 for stroke patients was 2.7%.at the national level, but it ranged from -12% to 24% across regions (the nine United States Census Bureau divisions). This varied response raises concerns that the hospital discharge process may be driven by payment policy rather than by clinical needs and individual preferences. Furthermore, varied changes in PAC use across regions might lead to untoward consequences, such as early hospital readmission. As efforts continue to reform PAC services and payment systems, it is essential that policymakers understand how different payment mechanisms associate with geographic variation in PAC use. The proposed study seeks to: 1) analyze geographic variation in PAC use before and after the BBA changes; 2) explore whether utilization and cost have shifted among PAC settings and whether early hospital readmission has increased; and 3) investigate how the contributions of patient, hospital, and market area characteristics in explaining PAC use differ between pre- and post-BBA periods. We will analyze the Center for Medicare & Medicaid Services' 5% sample of Medicare claims data from 1996 to 2000 to study the initial effect of the BBA changes on geographic variation in PAC use. We will focus on six diseases associated with high PAC use: stroke, hip procedure, hip fracture, chronic obstructive pulmonary disease, pneumonia, and congestive heart failure. The selected diseases provide a contrast between rehabilitative and medical conditions. The stability, the degree, and the association of geographic variation in PAC use before and after the BBA changes will be examined. Shifts in utilization and costs will be presented as correlations between changes in PAC use, hospital length of stay, and early hospital readmission. Finally, we will estimate multinomial logit models to explore changes in contribution to explain PAC use by patient, hospital, and market area characteristics after the BBA changes.